review use of characteristic of part of

[sbello]

We should review use of characteristic of part of vs characteristic of in eqs

From Slack chat with Nico

Nico:
why is http://purl.obolibrary.org/obo/upheno/patterns-dev/abnormalConstrictionOfAnatomicalEntity.yaml using characteristic of part of? Can we change that to “characteristic of”?
We had sort of settled on the general pattern that characteristic of part of should only be used for very general, high level patterns..

Me: I do not recall restricting the use of 'characteristic of part of' to only high level terms. My understanding is that we should use 'characteristic of part of' whenever we want to include all subparts of the entity as child terms so abnormal heart morphology uses 'characteristic of part of' and so does abnormal mitral valve morphology or abnormal mitochondria morphology but increased heart size does not as we would not want to say that increased mitral valve size is a increased heart size.
I think we had decided that we wanted the constricted pattern to include the subparts but we could review this if anyone thinks it shouldn't.

Nico:
lets say we have a constricted aorta. Do you want a constricted ion channel in a heart cell be a subclass of that?

Me: We wouldn't want that but wouldn't the ion channel be using a different pattern since the ion channel would not be an anatomical entity but instead a cellular component. We'd want to go down the anatomical entity chain but not cross over into the cellular components. Would using the characteristic of part of in the anatomy pattern automatically cross over the patterns?

Nico: I think (but not 100% sure) that go nucleus is part of cl cell. that means yes, it would indeed cross the boundaries.

Assuming that the part of crosses boundaries between ontologies based on relations between ontologies (cardiac ion channel part of heart cell part of heart). Then we should be more conservative in the usage of part of.

[cmungall]

If you want to infer this rigorously we need consistent section_part_of axioms in uberon. constricting a section of a tube in a tube/tree system also constricts the supersections. Otherwise you open up to many possible false positives. It so happens that with the current ontology structures we have we wouldn't see many of those false positives manifest, since we are reasoning on the class level and existential restrictions are conservative. But you get the false inference @matentzn mentions on the instance level.

Most of the existential descendants of aorta in Uberon and AOs so happen to be subsections, with the others being walls, and you generally wouldn't talk of a wall being restricted, so you're unlikely to see FPs manifest at the TBox level.

But the TL;DR is it's best to be safe and conservative, this means asserting knowledge as GCIs in phenotype ontologies, possibly duplicatively...